←back to thread

NoProp: Training neural networks without back-propagation or forward-propagation

(arxiv.org)
161 points belleville | 1 comments | 14 Apr 25 00:03 UTC | HN request time: 0.527s | source
Show context
DrFalkyn ◴[14 Apr 25 05:27 UTC] No.43678328[source]▶
If we could ever figure out what wet brains actually do (continuous feedback, ? enzyme release ? ) this might be possible
replies(2): >>43679873 #>>43680009 #
tsimionescu ◴[14 Apr 25 10:40 UTC] No.43679873[source]▶
Keep in mind that our brains also have a great deal of built in trained structure from evolution. So even if we understood exactly how a brain learns, we may still not be able to replicate it if we can't figure out the highly optimized initial state from which it starts in a fetus.
replies(4): >>43679958 #>>43680014 #>>43680557 #>>43681865 #
1. friendzis ◴[14 Apr 25 11:06 UTC] No.43680014[source]▶
I concur. It might not be feasible in terms of computational power available, but I don't think there is anything fundamentally stopping application of those training mechanisms, unless the whole neuralnet paradigm is fundamentally incompatible with those learning methods.